In partnership with

Today, let's explore "vibe coding" – a phenomenon that's reshaping how designers and non-technical creators can build functional digital experiences.

The intersection of AI and design is evolving rapidly, introducing new methodologies that challenge our traditional understanding of direct control of our creative acts to design, code and other creative activities.

There are some mindset shifts that can help such as thinking of the agents as your technical collaborator who you need to give direction to realize your vision. This helps clarify the types of user stories, wireframes and product requirement document to help realize your vision.

Control and Direction Are Not The Same

Inherent in the concept of vibing with our AI agents is to the idea of ‘collaboration’ which is an act of working together. Human AI collaborations means we cannot bring preconceived notions of realizing exactly what is in our head. Just as a human collaborator will bring their own creativity and insight to a task, so to will AI agents that are vibe coding with us. Sometimes this act of collaboration has the capacity to surprise and disappoint us. It is up to us art directors to provide the necessary detail to guide and determine when a deliverable meets the requirements and is done.

The Control Dilemma

When you vibe code, you're essentially delegating technical implementation to an AI system that may make architectural decisions you don't fully understand. This creates several potential issues:

• Verification challenges: Without technical knowledge, how can you verify the code works as intended beyond surface-level testing?

• Maintenance concerns: What happens when something breaks and you can't diagnose the problem?

• Dependency risk: You're relinquishing control to systems that may change their capabilities or pricing models.

• Test design: To effectively design unit tests you should understand the architecture and capabilities of the systems you create.

The Management Paradox

It's akin to managing a team without management experience – you can direct outcomes but lack insight into processes. This disconnect can lead to inefficient solutions or technical debt that's invisible to the vibe coder. It can create cybersecurity gaps and safety risks to the user if dangerous information or inaccurate data were to be released.

Optimists would argue that these systems are evolving at a rapid pace and becoming capable of addressing and co-creating solutions to many of the concerns human managers will raise. Further, AI agents have become increasingly capable of critiquing and diagnosing and then trouble-shooting issues in their own work. This capacity to diagnose and repair will only improve as multi-agent systems become the norm over time.

Regaining Control: Bridging the Knowledge Gap

Today to learn about a coding framework developers much search StackOverflow for answers to fix bugs and learn new techniques, AI agents allow for iterative learning through question and answer about the codebase:

• "Explain how the state management works in this code"

• "What are the security implications of the authentication approach you've implemented?"

• "Can you walk me through how the data flows between components?"

Non-technical human users can be coached on how to approach a technical safety inquiry to ensure robust and transparent quality assurance process.

Resources to Elevate Your Vibe Coding Practice

Integrated Development Environments (IDEs)

Advanced coding environments represent the cutting edge of development tools, offering everything from code completion to version control and error handling dramatically streamlining the development process. The most popular IDEs I’ve tested are:

Windsurf Editor represents the best choice for development teams focused on maximizing productivity through advanced AI assistance. Development teams working on complex applications who value cutting-edge AI assistance would benefit most from Windsurf's capabilities.

V0 by Vercel excels at democratizing web development, making it accessible to team members beyond traditional developers. Its seamless integration with Vercel's deployment platform creates a frictionless path from development to production for web applications. Organizations seeking to involve non-technical stakeholders in the development process, particularly for Next.js and React-based web projects.

Cursor is a popular option for enterprise development and has gained significant traction in the enterprise space, with 53% of Fortune 1000 companies having engineers using Cursor and over 30,000 enterprises choosing it as their IDE of choice. It's designed to handle large-scale development projects and offers features tailored for enterprise needs.

Replit offers the most comprehensive solution for collaborative development, making it approachable for both beginners to enterprise contexts. It is capable of building and deploying end-to-end similar to Vercel and Cursor.

Lovable Dev excels at rapid prototyping and development using natural language inputs, making it ideal for startups or individuals needing to quickly validate ideas or build MVPs.

Bolt is excellent for rapid prototyping and simple web applications, particularly for individuals and beginners.

LLMs including Claude, ChatGPT, Gemini and Grok can write apps too. They have the capacity to support writing the requirements documents and then write the code for the web app. Claude is even capable of publishing a widget. The difference I see in working with an LLM is the interface making editing and version control more challenging than with a specialized IDE like Replit.

A Critique of the Vibe Coding Experience

The benefit of lowering the bar to development is we can afford to make hyper specialized applications for individuals, families or small teams. Product market fit becomes an act of self reflection as to whether an applications serves my need or needs further refinement.

After creating this family chores application through vibe coding, several observations emerge:

Strengths:

• Rapid development: The entire application took minutes to conceptualize and generate

• Personalization: The solution addresses a specific family need rather than fitting into a generic productivity tool

• Accessibility: Created without writing a single line of code manually

• Design coherence: Maintains consistent visual language and interaction patterns

• Hyper specialization: anyone is able to make an app tailored to them, not some persona they must create product market fit around, but just them specifically.

Limitations:

• Customization ceiling: Some specialized interactions required multiple iterations to get right

• Structural understanding: The underlying component architecture isn't immediately apparent

• Data concerns: As there is no responsible organization management of my cybersecurity is in question. I wouldn’t use this approach for sensitive info.

Moving Beyond Coding To Other "Vibe" Activities

Let’s consider vibe coding, not as a one-off capability, but rather a principle for how we might approach our relationship to working with AI agents. What are other forms of vibing that the future will hold?

• Vibe designing: Creating visual assets through description rather than manual tool manipulation.

• Vibe writing: Collaborative content creation where AI expands on conceptual outlines from the user.

• Vibe data analysis: Extracting insights from datasets through conversational queries rather than writing analytical code

What are other activities where you would be comfortable holistically driving a capable AI agent team? Consider what would need to be true for you to trust the output just as you trust the output of a capable human associate.

The Future of Creation Through Conversation

Vibe coding represents more than a temporary shortcut – it's potentially a fundamental shift in how we approach software creation. While it won't replace traditional development for complex systems yet, it opens new creative possibilities for designers, entrepreneurs, and anyone with ideas but without traditional coding expertise.

Today, these AI coding agents are the worst they will ever be; stay ahead of the curve and begin experimenting with them now and imagining where they will be in the not too distant future with capabilities to produce secure, safe, reliable, scalable solutions to your specific challenges.

The question becomes not whether you can write perfect code, but whether you can clearly articulate your vision and guide an AI partner through the creation process. The importance of knowing the technical language for design and development to be able to clearly ask for specific solutions from UI style to data structure.

What are your experiences with vibe coding? Have you discovered effective prompts or methodologies to share with the community?

Hire an AI BDR to Automate Your LinkedIn Outreach

Sales reps are wasting time on manual LinkedIn outreach. Our AI BDR Ava fully automates personalized LinkedIn outreach using your team’s profiles—getting you leads on autopilot.

She operates within the Artisan platform, which consolidates every tool you need for outbound:

• 300M+ High-Quality B2B Prospects

• Automated Lead Enrichment With 10+ Data Sources Included

• Full Email Deliverability Management

• Personalization Waterfall using LinkedIn, Twitter, Web Scraping & More

Book a demo to see what Ava can do.

🧲 AI Agent Magnet

The video by Andrej Karpathy is a guide to using Large Language Models (LLMs) like ChatGPT in practical ways. It offers examples for various settings and applications, and it shows different LLM options, highlighting the differences between the original ChatGPT and its competitors such as Google's Gemini and Anthropic's Claude. Karpathy explains how these LLMs work, including their pre-training and post-training stages, and discusses the concept of a "thinking model," which uses reinforcement learning to improve accuracy in complex tasks. The video also explores tool use, such as enabling internet search and using a Python interpreter, alongside multi-modal interaction through audio and images. 

Spreadsheets are the default custom software generations of professionals have used, with the advent of vibe coding we can begin to replace these spreadsheets with custom built software to fit the specific needs of a company, team and even individual user. This post felt relevant as we discuss vibe coding and what it means to replace some of our spreadsheets with custom vibe coded tailor-made software solutions.

This is an interesting thread with some interesting starting-points for writing thin, light-weight product specification prompts to help you vibe code with an AI agent. Just like creating collateral for human collaboration writing user stories or a product requirements document can help articulate your vision clearly to your AI agents and what the goal of the final program.

💬 Suggestion box

A newsletter exploring the principles that will help us design AI agent experiences and startups that amplify human creativity.

Subscribe to join a community of designers and developers shaping purposeful AI agents.

New Referral Rewards

Until next time, keep innovating and stay curious!

Jason